brlock_ctdb.c

samba最新软件

/* 
   Unix SMB/CIFS implementation.

   generic byte range locking code - ctdb backend

   Copyright (C) Andrew Tridgell 2006
   
   This program is free software; you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
   the Free Software Foundation; either version 3 of the License, or
   (at your option) any later version.
   
   This program is distributed in the hope that it will be useful,
   but WITHOUT ANY WARRANTY; without even the implied warranty of
   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
   GNU General Public License for more details.
   
   You should have received a copy of the GNU General Public License
   along with this program.  If not, see <http://www.gnu.org/licenses/>.
*/

#include "includes.h"
#include "system/filesys.h"
#include "lib/tdb/include/tdb.h"
#include "messaging/messaging.h"
#include "lib/messaging/irpc.h"
#include "libcli/libcli.h"
#include "cluster/cluster.h"
#include "ntvfs/ntvfs.h"
#include "ntvfs/common/brlock.h"
#include "include/ctdb.h"

enum my_functions {FUNC_BRL_LOCK=1, FUNC_BRL_UNLOCK=2, 
		   FUNC_BRL_REMOVE_PENDING=3, FUNC_BRL_LOCKTEST=4,
		   FUNC_BRL_CLOSE=5};

/*
  in this module a "DATA_BLOB *file_key" is a blob that uniquely identifies
  a file. For a local posix filesystem this will usually be a combination
  of the device and inode numbers of the file, but it can be anything 
  that uniquely idetifies a file for locking purposes, as long
  as it is applied consistently.
*/

/* this struct is typically attached to tcon */
struct brl_context {
	struct ctdb_context *ctdb;
	struct ctdb_db_context *ctdb_db;
	struct server_id server;
	struct messaging_context *messaging_ctx;
};

/*
  the lock context contains the elements that define whether one
  lock is the same as another lock
*/
struct lock_context {
	struct server_id server;
	uint16_t smbpid;
	struct brl_context *ctx;
};

/* The data in brlock records is an unsorted linear array of these
   records.  It is unnecessary to store the count as tdb provides the
   size of the record */
struct lock_struct {
	struct lock_context context;
	struct ntvfs_handle *ntvfs;
	uint64_t start;
	uint64_t size;
	enum brl_type lock_type;
	void *notify_ptr;
};

/* this struct is attached to on open file handle */
struct brl_handle {
	DATA_BLOB key;
	struct ntvfs_handle *ntvfs;
	struct lock_struct last_lock;
};

#if 0
static void show_locks(const char *op, struct lock_struct *locks, int count)
{
	int i;
	DEBUG(0,("OP: %s\n", op));
	if (locks == NULL) return;
	for (i=0;i<count;i++) {
		DEBUG(0,("%2d: %4d %4d %d.%d.%d %p %p\n",
			 i, (int)locks[i].start, (int)locks[i].size, 
			 locks[i].context.server.node,
			 locks[i].context.server.id,
			 locks[i].context.smbpid,
			 locks[i].context.ctx,
			 locks[i].ntvfs));
	}
}
#endif

/*
  Open up the brlock.tdb database. Close it down using
  talloc_free(). We need the messaging_ctx to allow for
  pending lock notifications.
*/
static struct brl_context *brl_ctdb_init(TALLOC_CTX *mem_ctx, struct server_id server, struct loadparm_context *lp_ctx,
				    struct messaging_context *messaging_ctx)
{
	struct ctdb_context *ctdb = talloc_get_type(cluster_backend_handle(), 
						    struct ctdb_context);
	struct brl_context *brl;

	brl = talloc(mem_ctx, struct brl_context);
	if (brl == NULL) {
		return NULL;
	}

	brl->ctdb = ctdb;
	brl->ctdb_db = ctdb_db_handle(ctdb, "brlock");
	if (brl->ctdb_db == NULL) {
		DEBUG(0,("Failed to get attached ctdb db handle for brlock\n"));
		talloc_free(brl);
		return NULL;
	}
	brl->server = server;
	brl->messaging_ctx = messaging_ctx;

	return brl;
}

static struct brl_handle *brl_ctdb_create_handle(TALLOC_CTX *mem_ctx, struct ntvfs_handle *ntvfs, 
						    DATA_BLOB *file_key)
{
	struct brl_handle *brlh;

	brlh = talloc(mem_ctx, struct brl_handle);
	if (brlh == NULL) {
		return NULL;
	}

	brlh->key = *file_key;
	brlh->ntvfs = ntvfs;
	ZERO_STRUCT(brlh->last_lock);

	return brlh;
}

/*
  see if two locking contexts are equal
*/
static bool brl_ctdb_same_context(struct lock_context *ctx1, struct lock_context *ctx2)
{
	return (cluster_id_equal(&ctx1->server, &ctx2->server) &&
		ctx1->smbpid == ctx2->smbpid &&
		ctx1->ctx == ctx2->ctx);
}

/*
  see if lck1 and lck2 overlap
*/
static bool brl_ctdb_overlap(struct lock_struct *lck1, 
			struct lock_struct *lck2)
{
	/* this extra check is not redundent - it copes with locks
	   that go beyond the end of 64 bit file space */
	if (lck1->size != 0 &&
	    lck1->start == lck2->start &&
	    lck1->size == lck2->size) {
		return true;
	}
	    
	if (lck1->start >= (lck2->start+lck2->size) ||
	    lck2->start >= (lck1->start+lck1->size)) {
		return false;
	}
	return true;
} 

/*
 See if lock2 can be added when lock1 is in place.
*/
static bool brl_ctdb_conflict(struct lock_struct *lck1, 
			 struct lock_struct *lck2)
{
	/* pending locks don't conflict with anything */
	if (lck1->lock_type >= PENDING_READ_LOCK ||
	    lck2->lock_type >= PENDING_READ_LOCK) {
		return false;
	}

	if (lck1->lock_type == READ_LOCK && lck2->lock_type == READ_LOCK) {
		return false;
	}

	if (brl_ctdb_same_context(&lck1->context, &lck2->context) &&
	    lck2->lock_type == READ_LOCK && lck1->ntvfs == lck2->ntvfs) {
		return false;
	}

	return brl_ctdb_overlap(lck1, lck2);
} 


/*
 Check to see if this lock conflicts, but ignore our own locks on the
 same fnum only.
*/
static bool brl_ctdb_conflict_other(struct lock_struct *lck1, struct lock_struct *lck2)
{
	/* pending locks don't conflict with anything */
	if (lck1->lock_type >= PENDING_READ_LOCK ||
	    lck2->lock_type >= PENDING_READ_LOCK) {
		return false;
	}

	if (lck1->lock_type == READ_LOCK && lck2->lock_type == READ_LOCK) 
		return false;

	/*
	 * note that incoming write calls conflict with existing READ
	 * locks even if the context is the same. JRA. See LOCKTEST7
	 * in smbtorture.
	 */
	if (brl_ctdb_same_context(&lck1->context, &lck2->context) &&
	    lck1->ntvfs == lck2->ntvfs &&
	    (lck2->lock_type == READ_LOCK || lck1->lock_type == WRITE_LOCK)) {
		return false;
	}

	return brl_ctdb_overlap(lck1, lck2);
} 


/*
  amazingly enough, w2k3 "remembers" whether the last lock failure
  is the same as this one and changes its error code. I wonder if any
  app depends on this?
*/
static NTSTATUS brl_ctdb_lock_failed(struct brl_handle *brlh, struct lock_struct *lock)
{
	/*
	 * this function is only called for non pending lock!
	 */

	/* in SMB2 mode always return NT_STATUS_LOCK_NOT_GRANTED! */
	if (lock->ntvfs->ctx->protocol == PROTOCOL_SMB2) {
		return NT_STATUS_LOCK_NOT_GRANTED;
	}

	/* 
	 * if the notify_ptr is non NULL,
	 * it means that we're at the end of a pending lock
	 * and the real lock is requested after the timeout went by
	 * In this case we need to remember the last_lock and always
	 * give FILE_LOCK_CONFLICT
	 */
	if (lock->notify_ptr) {
		brlh->last_lock = *lock;
		return NT_STATUS_FILE_LOCK_CONFLICT;
	}

	/* 
	 * amazing the little things you learn with a test
	 * suite. Locks beyond this offset (as a 64 bit
	 * number!) always generate the conflict error code,
	 * unless the top bit is set
	 */
	if (lock->start >= 0xEF000000 && (lock->start >> 63) == 0) {
		brlh->last_lock = *lock;
		return NT_STATUS_FILE_LOCK_CONFLICT;
	}

	/*
	 * if the current lock matches the last failed lock on the file handle
	 * and starts at the same offset, then FILE_LOCK_CONFLICT should be returned
	 */
	if (cluster_id_equal(&lock->context.server, &brlh->last_lock.context.server) &&
	    lock->context.ctx == brlh->last_lock.context.ctx &&
	    lock->ntvfs == brlh->last_lock.ntvfs &&
	    lock->start == brlh->last_lock.start) {
		return NT_STATUS_FILE_LOCK_CONFLICT;
	}

	brlh->last_lock = *lock;
	return NT_STATUS_LOCK_NOT_GRANTED;
}

struct ctdb_lock_req {
	uint16_t smbpid;
	uint64_t start;
	uint64_t size;
	enum brl_type lock_type;
	void *notify_ptr;
	struct server_id server;
	struct brl_context *brl;
	struct ntvfs_handle *ntvfs;
};

/*
  ctdb call handling brl_lock()
*/
static int brl_ctdb_lock_func(struct ctdb_call_info *call)
{
	struct ctdb_lock_req *req = (struct ctdb_lock_req *)call->call_data->dptr;
	TDB_DATA dbuf;
	int count=0, i;
	struct lock_struct lock, *locks=NULL;
	NTSTATUS status = NT_STATUS_OK;

	/* if this is a pending lock, then with the chainlock held we
	   try to get the real lock. If we succeed then we don't need
	   to make it pending. This prevents a possible race condition
	   where the pending lock gets created after the lock that is
	   preventing the real lock gets removed */
	if (req->lock_type >= PENDING_READ_LOCK) {
		enum brl_type lock_type = req->lock_type;
		req->lock_type = (req->lock_type==PENDING_READ_LOCK? READ_LOCK : WRITE_LOCK);
		if (brl_ctdb_lock_func(call) == 0 && call->status == NT_STATUS_V(NT_STATUS_OK)) {
			return 0;
		}
		req->lock_type = lock_type;
	}

	dbuf = call->record_data;

	ZERO_STRUCT(lock);
	lock.context.smbpid = req->smbpid;
	lock.context.server = req->server;
	lock.context.ctx = req->brl;
	lock.ntvfs = req->ntvfs;
	lock.start = req->start;
	lock.size = req->size;
	lock.lock_type = req->lock_type;
	lock.notify_ptr = req->notify_ptr;

	if (dbuf.dptr) {
		/* there are existing locks - make sure they don't conflict */
		locks = (struct lock_struct *)dbuf.dptr;
		count = dbuf.dsize / sizeof(*locks);

		for (i=0; i<count; i++) {
			if (brl_ctdb_conflict(&locks[i], &lock)) {
				status = NT_STATUS_LOCK_NOT_GRANTED;
				goto reply;
			}
		}
	}

	call->new_data = talloc(call, TDB_DATA);
	if (call->new_data == NULL) {
		return CTDB_ERR_NOMEM;
	}

	call->new_data->dptr = talloc_size(call, dbuf.dsize + sizeof(lock));
	if (call->new_data->dptr == NULL) {
		return CTDB_ERR_NOMEM;
	}
	memcpy(call->new_data->dptr, locks, dbuf.dsize);
	memcpy(call->new_data->dptr+dbuf.dsize, &lock, sizeof(lock));
	call->new_data->dsize = dbuf.dsize + sizeof(lock);

	if (req->lock_type >= PENDING_READ_LOCK) {
		status = NT_STATUS_LOCK_NOT_GRANTED;
	}

reply:
	call->status = NT_STATUS_V(status);

	return 0;
}


/*
  Lock a range of bytes.  The lock_type can be a PENDING_*_LOCK, in
  which case a real lock is first tried, and if that fails then a
  pending lock is created. When the pending lock is triggered (by
  someone else closing an overlapping lock range) a messaging
  notification is sent, identified by the notify_ptr
*/
static NTSTATUS brl_ctdb_lock(struct brl_context *brl,
			      struct brl_handle *brlh,
			      uint16_t smbpid,
			      uint64_t start, uint64_t size, 
			      enum brl_type lock_type,
			      void *notify_ptr)
{
	struct ctdb_lock_req req;
	struct ctdb_call call;
	int ret;
	NTSTATUS status;

	call.call_id = FUNC_BRL_LOCK;
	call.key.dptr = brlh->key.data;
	call.key.dsize = brlh->key.length;
	call.call_data.dptr = (uint8_t *)&req;
	call.call_data.dsize = sizeof(req);
	call.flags = 0;
	call.status = 0;

	ZERO_STRUCT(req);
	req.smbpid = smbpid;
	req.start  = start;
	req.size   = size;
	req.lock_type = lock_type;
	req.notify_ptr = notify_ptr;
	req.server = brl->server;
	req.brl = brl;
	req.ntvfs = brlh->ntvfs;

	ret = ctdb_call(brl->ctdb_db, &call);
	if (ret == -1) {
		return NT_STATUS_INTERNAL_DB_CORRUPTION;
	}

	status = NT_STATUS(call.status);

	if (NT_STATUS_EQUAL(status, NT_STATUS_LOCK_NOT_GRANTED)) {
		struct lock_struct lock;
		lock.context.smbpid = smbpid;
		lock.context.server = brl->server;
		lock.context.ctx = brl;
		lock.ntvfs = brlh->ntvfs;
		lock.start = start;
		lock.size = size;
		lock.lock_type = lock_type;
		lock.notify_ptr = notify_ptr;
		status = brl_ctdb_lock_failed(brlh, &lock);
	}

	return status;
}

/*
  we are removing a lock that might be holding up a pending lock. Scan
  for pending locks that cover this range and if we find any then
  notify the server that it should retry the lock. In this backend, we
  notify by sending the list of locks that need to be notified on back
  in the reply_data of the ctdb call. The caller then does the
  messaging for us. 
*/
static int brl_ctdb_notify_unlock(struct ctdb_call_info *call,
				  struct lock_struct *locks, int count, 
				   struct lock_struct *removed_lock)
{
	int i, last_notice;

	/* the last_notice logic is to prevent stampeding on a lock
	   range. It prevents us sending hundreds of notifies on the
	   same range of bytes. It doesn't prevent all possible
	   stampedes, but it does prevent the most common problem */
	last_notice = -1;

	for (i=0;i<count;i++) {
		if (locks[i].lock_type >= PENDING_READ_LOCK &&
		    brl_ctdb_overlap(&locks[i], removed_lock)) {
			struct lock_struct *nlocks;
			int ncount;

			if (last_notice != -1 && brl_ctdb_overlap(&locks[i], &locks[last_notice])) {
				continue;
			}
			if (locks[i].lock_type == PENDING_WRITE_LOCK) {
				last_notice = i;
			}
			if (call->reply_data == NULL) {
				call->reply_data = talloc_zero(call, TDB_DATA);
				if (call->reply_data == NULL) {
					return CTDB_ERR_NOMEM;
				}
			}
			/* add to the list of pending locks to notify caller of */
			ncount = call->reply_data->dsize / sizeof(struct lock_struct);
			nlocks = talloc_realloc(call->reply_data, call->reply_data->dptr, 
						struct lock_struct, ncount + 1);